Publikationen - Netze für Erneuerbare Energien
|
Eberhard Waffenschmidt, Markus de Koster, Patrick Lehnen, In the near future electrical power grids need to supply a number of additional loads and decentralized generated renewable power. This requires an increased effort to control and supervise those dispatchable loads and components. Machine learning algorithms like neural networks can help significantly in these tasks. Examples, which will be presented, are predicting grid states with a drastically reduced effort of measurement equipment, analysing huge amount of measurements for irregularities or locating origins for disturbances in a power grid. Neural networks, which include and consider physical information, are especially well suited for such tasks. Downloads> Presentation: PDF-Dokument (0.97 MB) Thema |
 Principle of a physics aware neural network (PANN): The physical structure of the power grid is reflecetd in the topology of the neural network. |
E.Waffenschmidt, 14.Mai 2025